autobiographical memory reactivation in empathy: eeg and ...autobiographical memory reactivation in...

Autobiographical Memory reactivation in Empathy

1

1

2

EEG and fMRI Evidence for Autobiographical Memory Reactivation 3

in Empathy 4

Federica Meconi*1, Juan Linde-Domingo1,2, Catarina S. Ferreira1, Sebastian Michelmann1,3, 5

Bernhard Staresina1,4, Ian Apperly1, Simon Hanslmayr1,4,5 6

1 School of Psychology, University of Birmingham, B15 2TT. +44 12141 49517 7 2 Max Plank Institute Berlin for Human Development, 14195, Berlin; +49 30 82406-475 8

3 Princeton Neuroscience Institute, Princeton, NJ 08544. +1 (609) 258-0826 9

4 Center for Human Brain Health, University of Birmingham, B15 2TT. 10

5 Institute for Neuroscience and Psychology, University of Glasgow, G12 8QB 11

12

13

*Corresponding Author: Federica Meconi, [email protected] 14

15

16

Acknowledgements. 17

The Authors thank all the participants and Ross Wilson and Nina Salman for helping with fMRI data collection. 18

Funding: This work was supported by the European Union's Horizon 2020, MSCA-IF-2015 (Nº702530), and by 19

the ESRC (NºES/S001964/1) awarded to F.M. S.H. is supported by grants from the ERC (Nº647954), the ESRC 20

(NºES/R010072/1), and the Wolfson Society and Royal Society. 21

22

23

24

25

1Abbreviations 26

1 AM: Autobiographical memory; non-AM: non-autobiographical memory; LDA: Linear Discriminant Analysis.

IPL: Inferior Parietal Lobule; PCC: Posterior Cingulate Cortex; PHG: Parahippocampal Gyrus; SFG: Superior

Frontal Gyrus; SPL: Superior Parietal Lobule; TPJ: Temporoparietal Junction.

.CC-BY-NC-ND 4.0 International licenseavailable under anot certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprint (which wasthis version posted October 1, 2020. ; https://doi.org/10.1101/715276doi: bioRxiv preprint

mailto:[email protected]

https://doi.org/10.1101/715276

http://creativecommons.org/licenses/by-nc-nd/4.0/


2

27

Authors names and affiliations 28

Corresponding author: Federica Meconi (F.M.), School of Psychology, Hills Building, University of 29

Birmingham, B15 2TT. 30

Phone: +44 12141 49517; email address: [email protected] 31

Juan Linde-Domingo (J.L.D.), Max Plank Institute Berlin for Human Development, 14195, Berlin. 32

Phone: +49 30 82406-475; email address: [email protected] 33

Catarina S. Ferreira (C.S.F.), School of Psychology, Hills Building, University of Birmingham, B15 2TT. 34

Phone: +44 12141 49517; email address: [email protected] 35

Sebastian Michelmann (S.M.), Princeton Neuroscience Institute, Princeton, NJ 08544. 36

Phone: +1 (609) 258-0826; email address: [email protected] 37

Bernhard Staresina (B.S.), School of Psychology and Center for Human Brain Health, Hills Building, University 38

of Birmingham, B15 2TT. Phone: +44 12141 44913; email address: [email protected] 39

Ian A. Apperly (I.A.A.), School of Psychology, 52 Pritchatts Road, University of Birmingham, B15 2TT. Phone: 40

+44 12141 43339; email address: [email protected] 41

Simon Hanslmayr (S.H.), Institute for Neuroscience and Psychology, 62 Hillhead Street, University of Glasgow, 42

G12 8QB. email address: [email protected] 43

44

45

46

47










https://doi.org/10.1101/715276



3

Abstract 48

Empathy relies on the ability to mirror and to explicitly infer others’ inner states. Theoretical accounts suggest 49

that memories play a role in empathy but direct evidence of a reactivation of autobiographical memories (AM) in 50

empathy is yet to be shown. We addressed this question in two experiments. In experiment 1, electrophysiological 51

activity (EEG) was recorded from 28 participants who performed an empathy task in which targets for empathy 52

were depicted in contexts for which participants either did or did not have an AM, followed by a task that explicitly 53

required memory retrieval of the AM and non-AM contexts. The retrieval task was implemented to extract the 54

neural fingerprints of AM and non-AM contexts, which were then used to probe data from the empathy task. An 55

EEG pattern classifier was trained and tested across tasks and showed evidence for AM reactivation when 56

participants were preparing their judgement in the empathy task. Participants self-reported higher empathy for 57

people depicted in situations they had experienced themselves as compared to situations they had not experienced. 58

A second independent fMRI experiment replicated this behavioural finding and showed the predicted activation 59

in the brain networks underlying both AM retrieval and empathy: precuneus, posterior parietal cortex, superior 60

and inferior parietal lobule and superior frontal gyrus. Together, our study reports behavioural, 61

electrophysiological and fMRI evidence that robustly supports the involvement of AM reactivation in empathy. 62

Keywords: Empathy, Autobiographical memory, EEG, fMRI, EEG pattern classifier 63

64

65



https://doi.org/10.1101/715276



4

1. Introduction 66

When we encounter somebody who has a physical injury, like a broken leg, we feel we have a good 67

understanding of their pain, especially if we have experienced that same injury in our life. Therefore, it is 68

intuitively compelling to assume that empathy, i.e. the ability to share and understand others’ inner states, draws 69

on first-hand experiences we collected in our own past, i.e. on autobiographical memories (AM). However, this 70

compelling intuition cannot be taken for granted because empathy might instead be supported by semantic 71

memory about the general experience of pain and the conditions in which it is likely to occur (Perry et al., 2011; 72

Rabin and Rosenbaum, 2012). The present study used advanced imaging methods to distinguish carefully between 73

the roles of autobiographical and semantic memory, and seek the first direct evidence of reactivation of AM in 74

the service of empathy. 75

The claim for an interplay between AM and empathy is supported by several sources of convergent evidence. 76

Healthy students show higher empathy levels for adults experiencing chronic pain if they can rely on their own 77

general AM of physical pain when compared to control conditions (Bluck et al., 2013). Patients with congenital 78

insensitivity to pain show attenuated self-rated empathy for others’ pain, for which they could have collected no 79

experience or memory (Danziger et al., 2009). Several studies, provide evidence of common brain networks for 80

AM retrieval and cognitive empathy (i.e. reasoning explicitly about mental states (Buckner and Carroll, 2007; 81

Spreng et al., 2008; Spreng and Grady, 2009), with these brain areas including precuneus, posterior cingulate 82

cortex (PCC), retrosplenial cortex, medial temporal lobe (MTL), temporoparietal junction (TPJ) and medial 83

prefrontal cortex (mPFC, BA 10). Neuropsychological studies on patients with different kinds of memory 84

impairments have reported generally convergent results about impoverished empathic abilities as measured by 85

neuropsychological assessments or self-report questionnaires. Empathy failures are symptoms of several 86

psychiatric disorders with long-term memory impairment, e.g. schizophrenia (Corcoran and Frith, 2003; Meconi 87

et al., 2016). They are observed in patients with Alzheimer’s disease (Moreau et al., 2013; Ramanan et al., 2017), 88

Korsakoff’s syndrome (Drost et al., 2019; Oosterman et al., 2011), Mild Cognitive Impairment (Moreau et al., 89

2015), Parkinson disease (Monetta et al., 2009; Pell et al., 2014; Xi et al., 2015), and Semantic dementia (Duval 90

et al., 2012). In healthy ageing the affective side of empathy seems to decrease with age (Chen et al., 2014; Duval 91

et al., 2011; Ze et al., 2014). However, conclusions from the latter studies must be treated with some caution 92

because patients with a diagnosis involving different kinds of memory loss might show reduced empathy as a 93

consequence of a global cognitive decline that impairs multiple executive functions that in turn show reduced self-94

reported empathy. 95

Curiously, the handful of studies on patients with amnesia, i.e. a memory disorder observed after focal 96

hippocampal cortices damage that impairs the ability to consciously access AM, showed that empathy seems to 97

be spared (Rosenbaum et al., 2007) or only mildly impaired (Beadle et al., 2013; Staniloiu et al., 2013), at least 98

for cognitive empathy (“cold reasoning” about mental states) rather than affective empathy (“hot simulation” of 99

other people’s states and experiences, Sessa et al., 2014b; Shamay-Tsoory et al., 2009). While studies of amnesia 100

are potentially powerful sources of evidence evaluating the role of memory in empathy, they do not provide 101

definitive evidence about the role of AMs. This is because the retrieval of AMs does not rely only on the 102

hippocampal cortices but is underpinned by a network of brain areas that involves the prefrontal cortex and parietal 103

areas including precuneus, posterior parietal cortex and the retrosplenial cortex (Boccia et al., 2019; Cabeza and 104

St Jacques, 2007; Cotelli et al., 2012). In line with the systems consolidation account, by which memories become 105



https://doi.org/10.1101/715276



5

gradually independent of the hippocampus and stably stored in the neocortex, at least remote memories might still 106

be available as a source of semantic knowledge or implicit memories for the patients (Antony et al., 2017; 107

McClelland et al., 1995). This clearly leaves open the possibility that AMs are retrieved during empathy, and that 108

this could be detected with appropriate methods. 109

Even though the studies mentioned above support the idea that empathy draws on AM, critical evidence is 110

missing. In particular, it has not yet been demonstrated that AMs are actively retrieved in the service of empathy. 111

In order to test for evidence of a re-activation of AM when empathizing with others, we investigated healthy 112

adults’ empathy for an AM experience that the participants shared with the targets of empathy, in contrast with 113

an un-shared experience for which participants had no AM. In light of evidence that the role of AM may be 114

different for “affective” and “cognitive” empathy, we also varied whether the shared and un-shared experiences 115

were emotionally potent (physical pain) versus emotionally neutral (e.g., visiting a museum). 116

Recent advances in multivariate pattern analysis methods show that brain activity patterns can be tracked during 117

the encoding of new neutral episodes and re-observed during their retrieval (Linde-Domingo et al., 2019; 118

Michelmann et al., 2016). Furthermore, recent neuroimaging studies show that reinstatement of autobiographical 119

pain involves partial reinstatement of activity in the brain areas that process nociception (Fairhurst, Fairhurst, 120

Berna, & Tracey, 2012; Forkmann, Wiech, Sommer, & Bingel, 2015). Therefore, we here tested for a direct 121

evidence of online reactivation of AM when participants were required to explicitly rate their empathy awareness 122

for others’ neutral and painful expriences in two independent experiments (Fig. 1c, d). In experiment 1, EEG was 123

recorded during two sequential tasks: the first was a pain decision task, classically used to prompt an empathic 124

reaction, in which targets for empathy were depicted in contexts for which participants either did or did not have 125

an AM (Fig.1a); the second was a a task that explicitly required retrieval of the AM and non-AM contexts. The 126

retrieval task was used to extract the neural fingerprints of AMs and non-AMs (Fig.1b). A linear discriminant 127

analysis (LDA) EEG pattern classifier was trained during the retrieval task and tested on data obtained from the 128

preceding empathy task to test for the online reactivation of the memories in explicit empathy. In experiment 2, 129

fMRI was measured from an independent sample performing the same empathy task to test if it activated brain 130

areas commonly associated with empathy and AM. 131



https://doi.org/10.1101/715276



6

132

Fig. 1. Experimental Design a) Schematic representation of the empathy task used in experiment 1 and 133

experiment 2. Participants were required to rate how much empathy they felt for the person depicted in the 134

preceding context. b) Schematic representation of the retrieval task used in experiment1 that was used to train the 135

LDA classifier. Participants first learnt to associate four abstract figures with the same sentences describing 136

painful contexts presented during the empathy task (not shown here). In the actual task, for each trial participants 137

were presented with one of the four figures and had to picture in their mind’s eye the context that they learnt to 138

associate with that specific figure. c) Raincloud plots of the subjective reports of participants’ empathy awareness 139

in experiment 1 and experiment 2. d) Concept of the study; when we encounter someone who shares our same 140

physically painful experience, memory of that experience is reactivated to empathize. 141

2. Material and Methods 142

The protocol for both experiments was approved by the University of Birmingham Research Ethics 143

Committee (ERN-16-0101A). Written informed consent was obtained from all participants for both experiments. 144



https://doi.org/10.1101/715276



7

2.1. Participants 145

We aimed for a sample size of 28 participants. This is consistent with previous studies using the pattern classifier 146

in the field of memory (Linde-Domingo et al., 2019; Michelmann et al., 2016). Also, using the PANGEA analysis 147

tool (https://jakewestfall.shinyapps.io/pangea/), 28 participants was judged to be able to detect the main effect and 148

the 3-way interaction of a medium effect size with a power exceeding 0.9. All the participants, for both 149

experiments, were recruited through the Research Participant Scheme of University of Birmingham for cash 150

(£10/h) or course credits (1 credit/h). All of the participants had normal or corrected-to normal vision. The 151

eligibility criteria included native or excellent English proficiency, no history of neurological or psychiatric 152

disorder, and having an experience of intense physical pain in their past. In order to ascertain that all the eligibility 153

criteria were met, students who signed up for the study were contacted and screened before they were accepted as 154

participants for the studies. During this initial screening phase, students were asked about their English 155

proficiency, they had to complete a questionnaire were pathological history or psychotropic drugs assumptions 156

were checked and they had to complete the Autobiographical Memory Questionnaire, AMQ (Rubin et al., 2003), 157

for an experience of intense physical pain and for one emotionally and physically neutral that they were asked to 158

report. Students who could not report any experience of intense physical pain or of a neutral experience could not 159

be participants in these studies. 160

2.1.1. Experiment 1. EEG. 161

Thirty-five healthy students took part in the experiment (mean age = 22, SD = 5). Four were left-handed, 6 were 162

males. Seven were discarded from the final sample, three were not Caucasian (we showed pictures of Caucasian 163

faces and previous studies have shown that empathic responses are subject to ethnicity bias (e.g., Sheng & Han, 164

2012), two could not complete the task due to equipment failure and two for too low number of trials due to 165

inaccurate responses. The final sample was composed of 28 participants (mean age = 21.96, SD = 4.82), four were 166

males and four left-handed. 167

2.1.2. Experiment 2. fMRI. 168

Thirty three healthy students took part in the experiment (mean age = 25, SD = 5.9). Participants were all right-169

handed, 15 were males. Five were discarded from the final sample, two served as pilots to adjust the timing of the 170

paradigm and make it suitable for the fMRI environment; one participant could not complete the acquisition 171

session in the scanner, two were discarded for excessive movements (more than 1 voxel size, 3mm). The final 172

sample was composed of 28 participants (mean age = 24.71, SD = 5.86), eleven were males. 173

2.2. Questionnaires 174

As mentioned in section 2.1, before accepting students as participants for the study, the students who signed 175

up for the study underwent a screening phase. The screening consisted of collecting information about students’ 176

history of pathological morbididy, English proficiency and, critically, an AM of intense physical pain and of a 177

neutral experience, in terms of emotion and pain. Candidates were therefore asked to report these AMs and 178

complete the AMQ for both autobiographical episodes. The AMs reported by the participants were on average 179

4.65 years old (SD = 5 y) for experiment 1; 4.46 years old (SD = 5.72 y) for experiment 2. 180

At the end of the experimental session, dispositional empathy resources and the ability to recognize and 181

describe participants’ own emotions were assessed with the Empathy Quotient, EQ, and the Interpersonal 182

Reactivity Index, IRI (Baron-Cohen and Wheelwright, 2004; Davis, 1983) and with the Toronto Alexithymia 183

scale, TAS-20 (Bagby et al., 1994), respectively. Participants from both experiments fell in the normal range of 184



https://jakewestfall.shinyapps.io/pangea/

https://doi.org/10.1101/715276



8

the EQ (experiment 1: M = 51.14 SD = 9.99; experiment 2: M = 46.89 SD = 12.37), and had on average normal 185

ability to describe their emotions as showed by the TAS score (experiment 1: M = 45.96, SD = 12.22; Exp2: M = 186

43.96, SD = 12.01). The IRI scores for both experiments are reported in Table 1. These measures were also used 187

to explore any relation with the neural responses but no correlation was found significant (for further details on 188

the correlation analysis please see the Supplementary Materials). 189

190

Table 1. Questionnaires. Participants’ scores to the Interpersonal Reactivity Index in both experiments. T-tests 191

are performed on the two independent samples. 192

2.3. Stimuli and Procedure 193

All the stimuli were presented on a grey background of a 17’’ computer screen with a refreshing rate of 70 194

Hz. The tasks were programmed using Psychtoolbox. 195

2.3.1. Experiment 1: EEG 196

Participants performed two tasks in the same experimental session. For all participants, the first of these tasks was 197

the empathy task used in previous studies (Meconi et al., 2018; Sessa et al., 2014b) and the second one was a 198

retrieval task (Fig.1a and b). 199

2.3.1.1. The empathy task: stimuli and procedure 200

The stimuli for the empathy task were sentences, describing specific contexts, followed by faces, the task 201

required to rate participants’ empathy for the person as depicted in the preceding context. The faces were a set of 202

16 identities, 8 males and 8 females with a painful or a neutral facial expression. The faces were in shades of grey 203

and they were equalized for luminance with the SHINE toolbox (Willenbockel et al., 2010). The sentences 204

described contexts of a person feeling physical pain or depicted in an emotionally neutral context. The critical 205

manipulation in this task was that the targets of participants’ empathy were depicted in contexts for which they 206

had or had not a related AM. Therefore, two contexts (one describing physical pain and the neutral one) were 207

taken from participants’ autobiographical experience. In order to tailor the contexts for each participant, we 208

screened them prior to the experimental session as soon as they signed up for the study. Participants were asked 209

to report an experience of intense physical pain and an emotionally and physically neutral experience for which 210

they completed the AMQ. A physically painful and a neutral context that didn’t belong to participants’ 211

autobiographical experiences were also identified and used for the non-AM contexts. The four contexts identified 212

for each participant were described in the empathy task by a sentence that always followed the structure “This 213

person got – […]” or “This person did – […]” so that all the sentences had the same syntactic complexity. 214

Each trial started with a fixation cross (600 ms). Participants were then presented with the sequence of a 215

sentence (3 s) and a face (500 ms) interleaved by a variable fixation cross (800 -1600 ms jittered in steps of 100 216

ms). The task was to subjectively rate on a scale of 6 points how much empathy participants felt for the person as 217

depicted in the presented context. The rate was self-paced and presented after another fixation cross that was on 218

the screen for 500 ms. At the very end of the trial, participants were asked to indicate whether the face, regardless 219

of the context, had a painful or a neutral expression. The task was composed of 48 trials per condition that were 220



https://doi.org/10.1101/715276



9

pseudo-randomized in a way that the conditions were balanced over the full session. There were 192 trials in total 221

subdivided in 4 blocks. An illustration of the task is shown in Fig.1a. 222

2.3.1.2. The retrieval task: Stimuli and procedure 223

For the retrieval task, 16 shapes were created ad hoc in total. For each participant a unique subset of 4 of these 224

shapes was presented. The first step was to generate 8 random polygons with equal number of black pixels. The 225

polygons were then blurred with a Gaussian filter and all the pixels in shades of grey were made black to create 8 226

rounded shapes. As a last step, the number of black pixels was equalized across all the shapes (i.e., random 227

polygons and the rounded shapes). The shapes were shown on a grey background (Fig.1b). 228

In the retrieval task, participants were required to picture in their minds’ eye the contexts described in the 229

empathy task. This task acted for the EEG pattern classifier as a localizer to extract the neural fingerprints of the 230

contexts to then probe the data from the empathy task. Therefore, to avoid perceptual confounds when applying 231

the classifier across tasks, participants were presented with their unique subset of four shapes that were used to 232

cue the contexts described in the empathy task. Before starting the retrieval task, participants underwent a 233

“learning phase” in which they learnt to associate each shape with one of the contexts, each sentence-shape pair 234

was presented twice before memory for the associations was tested. The retrieval task could only start when full 235

accuracy was reached in the “learning phase”. To test memory for the sentence-shape association, participants 236

were presented only with a figure at a time and had to indicate what was the context associated with the shape. 237

One cue-word per sentence was chosen to cue to the related context and allow responses (e.g., we used “arm” as 238

a cue-word for the sentence “This person got their right arm broken”; “ligament” for “This person got their 239

ligament torn”; “Museum” for “This person visited the Birmingham Museum of Art” and “laptop” for “This 240

person bought a new laptop” and so forth). The four cue-words were placed equally spaced horizontally at the 241

center of the screen and their order was randomized with a Latin square in such a way that each word had the 242

same likelihood to appear at one of the four locations (e.g., “arm” “ligament” “museum” “laptop”; “museum” 243

“arm” “laptop” “ligament” etc.). Participants could press one of four keys on the computer keyboard that spatially 244

corresponded to the location of the cue-word (“d” for the cue-word appearing on the very left, “f” for the cue-245

word appearing the central left, “j” for the one at the central right and “k” for the one at the very right location). 246

The memory test could end after 8 correct answers, i.e. 2 times each pair. One error within a block of 8 trials 247

would be followed by the repetition of a new block of 8 trials, until 100% accuracy would be reached. Once the 248

memory associations test was successful participants could start the practice session of the retrieval task. 249

Participants could familiarize with the retrieval task with a block of 8 trials that could be repeated until they felt 250

confident they understood the task. 251

In the retrieval task, participants were only shown the shapes. In each trial, one shape was shown for 3 secs 252

and participants had to picture in their mind’s eye the context associated with that shape. Within the time the shape 253

was on the screen, they were required to rate the vividness of the context as soon as they could picture it in their 254

mind’s eye, by pressing one of six response keys “s”, “d”, “f”, “j”, “k”, “l”, with “s” for “not vivid at all” to “l” 255

for “very vivid”. If they did not press any button, a “No Response” was recorded and the trial excluded from the 256

analysis. Participants were then asked to indicate which context they saw in their mind’s eye. They could answer 257

in the same way as they did for the memory association test with the further option that if they could not remember 258

what context was associated with that shape, they could press the space bar for “forgotten” and move to the next 259

trial. Responses were not time-pressured but only correct trials were included in the analysis. There were 60 trials 260



https://doi.org/10.1101/715276



10

per condition that were pseudo-randomized to balance the distribution of all the conditions over the total of 240 261

trials that constituted the full session. The task, depicted in Fig.1b, right panel, was subdivided in 4 blocks. 262

2.3.2. Experiment 2: fMRI 263

The screening phase, the questionnaires and the procedure were the same as those used in experiment 1 with 264

the exception of the necessary adjustments in the timing of the events applied to the empathy task in order to make 265

it suitable for the fMRI environment (for additional details, see Supplementary Material S1.3.1.). 266

2.4. Data acquisition and analysis 267


The EEG was recorded using a BioSemi Active-Two system from 128 Ag/AgCl active electrodes. The EEG 269

was re-referenced offline to the average reference. Three additional external electrodes were placed below the left 270

eye and on the lateral canthi of each eye to record vertical electroculogram (EOG). EEG and vertical EOG signals 271

were digitized at a sampling rate of 1024 Hz via ActiView recording software (BioSemi, Amsterdam, the 272

Netherlands). 273

EEG data was analyzed with MATLAB (©Mathworks, Munich, Germany) using the open-source FieldTrip 274

toolbox (http://fieldtrip.fcdonders.nl/) and in-house Matlab routines. 275

2.4.1.1. Preprocessing 276

2.4.1.1.1. The Empathy task 277

EEG data was first segmented into epochs of 2 s, starting 1 s before the onset of the face. The epoched data 278

was visually inspected to discard large artifacts from further analysis. Further preprocessing steps included 279

Independent Component Analysis for ocular artifacts correction and re-referencing to average reference. After 280

removing trials which were contaminated by eye and muscle artefacts, an average of 45 trials (range: 34-48) 281

remained for AM and 45 trials (range: 37-48) for non-AM condition. 282

2.4.1.1.2. The Retrieval task. 283

EEG data were first segmented into epochs of 4 s, starting 1 s before the onset of the cue. The epoched data 284

was visually inspected to discard large artifacts from further analysis. Further preprocessing steps included 285

Independent Component Analysis for ocular artifacts correction and re-referencing to average reference. After 286

removing trials which were contaminated by eye and muscle artefacts, an average of 51 trials (range: 42-60) 287

remained for AM and 50 trials (range: 38-58) for non-AM condition. 288

2.4.1.2. Event-related potentials (ERPs) analysis: the Empathy task 289

ERPs were time-locked to the onset of the face. We computed ERPs in response to painful and neutral faces. 290

To test any involvement of memory in the empathy task, we contrasted ERPs time-locked to the onset of the faces 291

reflecting the processing of the preceding context (AM vs. non-AM). To check whether there was any difference 292

between emotional content of the memories we also contrasted painful and neutral memories separately for AM 293

and non-AM. 294

2.4.1.3. Linear Discriminant Analysis EEG pattern classifier. 295

Linear Discriminant Analysis (LDA) is a multivariate pattern analysis method that finds a decision boundary 296

that allows distinguishing the pattern of brain activity associated with one category of stimuli from the pattern of 297

brain activity that is associated with another category of stimuli. This is based on specified features of the EEG 298

signal. It can then estimate with certain accuracy whether the pattern of brain activity in data that was not used to 299

find the decision boundary, is more similar to one or the other category of stimuli. 300



https://doi.org/10.1101/715276



11

In order to reduce unwanted noise and computational time, the signal was filtered between 0.1 and 40 Hz and 301

down sampled to 128Hz before classification with a baseline correction window of 500ms before the onset of the 302

stimuli. 303

The LDA was trained and tested on the EEG raw patterns (i.e. amplitude of the signal on each of the 128 304

electrodes), for each participant and at each time point and regularized with shrinkage (Blankertz et al., 2011). 305

To make sure that the output was not biased by the signal to noise ratio due to the different amount of trials, we 306

equalized the number of trials for AM and non-AM before train the classifier. 307

The classifier was trained on the raw signal (i.e. amplitude of EEG on each electrode) acquired while participants 308

were performing the retrieval task in the time-window including the presentation of the cue to detect systematic 309

differences between the EEG patterns reflecting the representation of AM and non-AM contexts. It was then tested 310

on the signal independently acquired while participants were performing the preceding empathy task in the time-311

window from the onset of the face until the rating was made. The aim of the LDA was to test for the online 312

reactivation of the memory in preparation of the explicit judgement of participants’ empathy awareness. 313

Before training and testing the LDA in two different datasets, we trained and tested the classifier on the retrieval 314

task during the presentation of the cue to show that the task was successful to act as a localizer of the representation 315

of the AM and non-AM contexts. A K-fold cross-validation procedure with 5 repetitions was used to train and 316

test the classifier. The output of this analysis is the accuracy with which the classifier could distinguish between 317

the two memory contexts for each time-point over all trials and electrodes. Therefore, the LDA reduces the data 318

into a single decoding time course per dataset. 319

2.4.1.4. Source Analysis 320

A standardized boundary element model was used for source modelling, which was derived from an averaged 321

T1-weighted MRI dataset (MNI, www.mni.mcgill.ca). That was used in combination with individual electrode 322

positions. Individual electrodes’ coordinates were logged with a Polhemus FASTRAK device (Colchester, 323

Vermont, USA) in combination with Brainstorm implemented in MATLAB 2014b (MathWorks). For three 324

participants the standard electrode coordinates were used due to technical problems during the experimental 325

session. 326

For source reconstruction, a time-domain adaptive spatial filtering linear constrained minimum variance (LCMV) 327

beamformer (Van Veen et al., 1997), as implemented in fieldtrip was applied. Source analysis was carried out for 328

the time-domain ERP components that revealed significant results on the scalp level. 329

2.4.1.5. Statistical analysis 330

2.4.1.5.1. Behaviour: the Empathy Task 331

Mean proportions of accurate responses given within +/-2.5 SD from the average reaction time of each 332

participant and mean proportions of the empathy awareness scores were computed for each condition and inserted 333

in two repeated measures ANOVAs with a 2 (Emotional memory: Painful vs. Neutral) x 2 (Memory: 334

Autobiographical vs. non-Autobiographical) x 2 (Facial expression: Painful vs. Neutral) as within-subject factors. 335

Bonferroni corrected paired-sample t-tests were conducted when appropriate to explore significant interactive 336

effects. Partial eta squared (ƞp2) are reported for completeness and transparency. Effect sizes are reported as eta 337

squared (ƞ2) calculated as the ratio between the sum of squares of each effect and the sum of the sum of squares 338

of all the effects and their errors, 95% confidence intervals (CI) of the mean differences between conditions are 339

reported in squared brackets. 340



http://www.mni.mcgill.ca/

https://doi.org/10.1101/715276



12

2.4.1.5.2. ERPs 341

Cluster-based permutation tests were performed over the whole scalp and over a 1sec time-window on the 342

event-related potentials time-locked to the onset of the face. We tested for significant differences between painful 343

and neutral facial expressions in order to replicate previous findings and show an ERP empathic response to faces. 344

Additionally, preliminary analysis was carried out to test for any involvement of memory in the pain decision task 345

and whether there was any difference related to the emotional content of the memory. To this end, cluster-based 346

permutation tests were performed on the ERPs time-locked to the onset of the face regardless of the facial 347

expression contrasting AM vs. non-AM and painful and neutral contexts separately for AM and non-AM. 348

2.4.1.5.3. LDA classifier 349

For the classifier analysis, an empirical null distribution was created with a combined permutation and 350

bootstrapping approach (Stelzer et al., 2013) that tested whether the maximum cluster of accuracy values above 351

the chance level was statistically significant. Clusters were identified on the basis of the number of adjacent pixels 352

found with the matlab function bwlabel. We used the LDA in 100 matrices with pseudo-randomly shuffled labels 353

independently for each participant and created a null distribution of accuracy values that we contrasted with the 354

LDA outputs obtained with the real data. This was done by sampling with replacement 100000 times from the 355

real and random data of each subject and computing a group average. This procedure resulted in an empirical 356

chance distribution, which allowed us to investigate whether the results from the real-labels classification had a 357

low probability of being obtained due to chance (p < 0.05) (i.e., exceeding the 95th percentile). 358

2.4.2. Experiment 2. fMRI 359

Data acquisition was performed with 3T Philips Medical Systems Achiva MRI scanner using a 32-channel 360

head coil. Functional T2-weighted images were acquired with isotropic voxels of 3mm, repetition time [TR] = 361

1750 msec, echo time [TE] = 30 msec, field of view [FOV] = 240x240x123 mm, and flip angle = 78°. Each 362

volume comprised 33 sequentially ascending axial slices with an interslice gap of 0.75mm). Each participant 363

underwent four blocks of scan series, one full block comprised 410 volumes. A high-resolution T1-weighted 364

anatomical scan was acquired with an MPRAGE sequence (TR = 7.4 msec, TE = 3.5 msec, isotropic voxel size 365

of 1mm, FOV = 256x256x176, flip angle = 7°) after the first two functional scanning blocks. The MR scanner 366

was allowed to reach a steady state by discarding the first three volumes in each of the four scan series block. 367

2.4.2.1. Preprocessing 368

The analyses were performed using the SPM12 toolbox (University College London, London, UK; 369

http://www.fil.ion.ucl.ac.uk/spm/). For each scanning block, a motion realignment of each slice to the first slice 370

was carried out before time realignment (slices corrected to the middle one). Data was then linearly detrended, 371

using a Linear Model of Global Signal algorithm (Macey et al., 2004) to remove any minimal fluctuation due to 372

the physical setting. Functional images served as reference for the co-registration with the anatomical image. The 373

data was further normalized to an MNI template, and finally, images were spatially smoothed with an 8-mm 374

FWHM Gaussian kernel. 375

2.4.2.2. Whole-brain analysis 376

Two separate univariate analyses were carried out for two different time-windows, one analysis was time-377

locked to the onset of the face, and the other was time-locked to the onset of the context. This was only done to 378

parallel experiment 1 and not to test for any functional dissociation between the two time-windows. In both cases 379

statistical parametric maps were created for each participant's block of trials. 380



https://doi.org/10.1101/715276



13

AM and non-AM conditions were directly contrasted in paired-sample t-tests on a group-level analysis. 381

The first analysed fMRI data were time-locked to the onset of the context. Regressors were defined for AM and 382

non-AM related to the onset of the contexts regardless of the emotional content of the context described by the 383

sentence. Additional regressors of no interest were again included in the design matrix to explain variance in the 384

data not due to the experimental manipulation under investigation and the 6 motion parameters obtained during 385

the realignment phase of the preprocessing. Sixty statistical parametric maps were created (4 blocks × 15 386

regressors) for each participant. 387

The second analysed fMRI data were time-locked to the onset of the face. Regressors were defined for 388

autobiographical and non-autobiographical memories time-locked to the onset of the faces regardless of their 389

emotional expression. Additional regressors of no interest were included in the design matrix to explain variance 390

in the data not due to the experimental manipulation under investigation plus the 6 motion parameters. Fifty-four 391

statistical parametric maps were created (4 blocks × 14 regressors) for each participant. Additional information 392

on the regressors of no interest are reported in Supplementary Material S1.3.2.2. 393

2.4.2.3. Statistical analysis 394

2.4.2.3.1. Behavior 395

Mean proportions of the empathy awareness scores were computed for each condition and inserted into a 396

repeated measures ANOVA with a 2 (Emotional memory: Painful vs. Neutral) x 2 (Memory: Autobiographical 397

vs. non-Autobiographical) x 2 (Facial expression: Painful vs. Neutral) as within-subject factors. Bonferroni 398

corrected paired-sample t-tests were conducted when appropriate to explore significant interactive effects. Partial 399

eta squared are reported for completeness and transparency. Effect sizes are reported as eta squared (ƞ2) calculated 400

as the ratio between the sum of squares of each effect and the sum of the sum of squares of all the effects and their 401

errors, 95% confidence intervals (CI) of the mean differences between conditions are reported in squared brackets. 402

2.4.2.3.2. fMRI 403

For both time-windows, a within-subject analysis was carried out on the data set of each participant to obtain 404

the mean statistical parametric map for each experimental condition. Finally, a group-level paired-sample t-test 405

contrasting AM and non-AM was performed. A cluster-wise analysis was performed with uncorrected p = .001 406

and then Family Wise Error correction was applied for multiple comparison (cluster p threshold = .05). Peak 407

voxel MNI are reported in brackets. Further information on the fMRI analysis and results can be found in the 408

Supplementary Materials (S1.3.2.1 and S2.2.1). 409

3. Results 410

3.1. Behavioural results 411


Individual scores of the empathy awareness revealed a main effect of the type of memory F(1,27) = 22.319; 413

p = .000064, ηp2 = .453, η2 = .092, Mdiff = .767 CI95 = [.434 1.10] such that AM context induced higher empathy 414

rates than non-AM contexts; of the emotional content of the memory F(1,27) = 50.902; p < .000001, ηp2 = .653, 415

η2 = .157, Mdiff = 1.0, CI95 = [-1.288 -.713] and of the facial expression F(1,27) = 42.270; p < .000001, ηp2 = .613, 416

η2 = .193, Mdiff = 1.108, CI95 = [-1.456 -.760] such that painful conditions induced higher rates than neutral 417

conditions. The two-way interaction between emotional content of the memory and of the facial expression was 418

significant F(1,27) = 18.390; p = .000206, ηp2 = .405, η2 = .08. Further exploration of the two-way interaction 419

revealed that painful faces drove higher rates of empathy awareness than neutral faces when the emotional content 420



https://doi.org/10.1101/715276



14

of the preceding memory was painful (t(27) = 9.7, pc < .0000001, Mdiff = 1.821, CI95 = [1.436 2.207]) but not 421

when it was neutral (pc = .167). In the same vein, painful memories reported higher empathy awareness scores 422

than neutral memories when followed by painful (t(27) = 10.825, pc < .0000001, Mdiff = 1.714, CI95 = [1.389 423

2.038]) but not neutral faces (pc = .286). The three-way interaction between the three factors was also significant 424

F(1,27) = 11.002; p = .003, ηp2 = .290, η2 = .003 such that empathy rates for painful faces was higher for painful 425

when compared to neutral contexts for both AM (t(27) = 8.219, pc < .0001, Mdiff = 1.927, CI95 = [1.165 2.689]) 426

and non-AM (t(27) = 6.397, pc < .0001, Mdiff = 1.500, CI95 = [.738 2.262]) but this difference was higher for AM 427

than non-AM contexts (t(27) = 2.122, p = .043, Mdiff = .427, CI95 = [.014 .840]). 428

3.1.2. Experiment 2: fMRI 429

Individual scores of the empathy awareness revealed a main effect of the type of memory F(1,27) = 7.210; p 430

= .012, ηp2 = .211, η2 = .021, Mdiff = .355, CI95 = [.084 .626]; of the emotional content of the memory F(1,27) = 431

48.860; p < .000001, ηp2 = .644, η2 = 0.186, Mdiff = 1.064, CI95 = [.752 1.376] and of the facial expression F(1,27) 432

= 38.863; p = .000001, ηp2 = .590, η2 = 0.186, Mdiff = 1.065, CI95 = [.714 1.415]. The two-way interaction between 433

emotional content of the memory and of the facial expression was significant F(1,27) = 19.995; p = .000126, ηp2 434

= .405, η2 = .096, so was the one between the emotional content and the type of memory F(1,27) = 4.758; p = 435

.038, ηp2 = .150, η2 = .008. Further exploration of the two-way interactions revealed that painful faces drove higher 436

rates of empathy awareness than neutral faces when the emotional content of the preceding memory was painful 437

(t(27) = 8.8, pc < .0000001, Mdiff = 1.83, CI95 = [1.404 2.257]) but not when it was neutral (pc = .280). In the same 438

vein, AM reported higher empathy awareness scores than non-AM when they were painful (t(27) = 3.016, pc = 439

.006, Mdiff = .578, CI95 = [.185 .971]) but not when they were neutral (pc = .348). However, painful memories 440

reported higher scores of empathy than neutral memories, were they either autobiographical or not (min t(27) = 441

5.17, pc = .00002, Mdiff = .841, CI95 = [.507 .1.175]). The three-way interaction did not reach significance level 442

F(1,27) = 2.294; p = .142, ηp2 = .078, η2 = .0006. 443

The behavioural results from the two experiments showed that individuals depicted in contexts describing 444

participants’ autobiographical contexts drove enhanced explicit judgements of empathy awareness when 445

compared to contexts describing non-autobiographical contexts independently of all the other factors. These 446

results are shown in Fig.1c). 447

3.2. EEG results 448

3.2.1. ERPs and source analysis 449

Cluster analysis conducted over a 1sec time-window, from the onset of the face until the presentation of the 450

rating, revealed one anterior and one posterior cluster of electrodes showing that ERPs significantly differ as a 451

function of the type of memory (anterior: p = 0.002; posterior: p = 0.002). Fig.2 depicts ERPs for AM and non-452

AM in the left panel and the topography of the significant clusters in right upper panel, t-values are plotted). 453

Source analysis estimated that the neural source of this effect was the Superior Frontal Gyrus, BA 10, MNI: [-10 454

69 0] (Fig.2 right bottom panel). No difference was found for separate contrasts between emotional contents of 455

the context for neither AM (p = .06) nor for non-AM (p = .18), therefore we did not further analyze differences 456

between emotional contents of the contexts. 457

458



https://doi.org/10.1101/715276



15

459

Fig. 2. ERPs results Left panel: ERPs time-locked to the onset of the face and reflecting AM and non-AM at the 460

anterior (upper panel) and the posterior cluster (bottom panel. Right top panel: clusters analysis performed over 461

all the electrodes in a 0 – 1 sec time-window. Colors code t-values. Right bottom panel: source localization of the 462

AM vs non-AM contrast. 463

Additionally, in line with previous studies on empathy for physical pain, cluster analysis also revealed a classic 464

ERP response associated with empathic processes (e.g., Sessa, Meconi, & Han, 2014), i.e. painful faces elicited 465

more positive ERP responses than neutral faces (p = 0.004). Consistently, source analysis estimated that the neural 466

source of this effect was the Inferior Frontal Gyrus, BA 9, MNI: [-62 21 30] and the Parietal Lobule, BA 7, MNI 467

[30 -69 48]. 468

3.2.2. LDA 469

We first ran a sanity check of the classifier on the retrieval task. The classifier was trained and tested with a 470

K-fold cross-validation procedure during the presentation of the cue (Fig.1b) as reported in section 2.4.1.3. Since 471

we were interested in investigating whether AM is reactivated in empathy, we checked that the classifier could 472

distinguish first of all whether the context pictured in the participants’ mind’s eye was an AM or a non-AM. 473

The square-shape of the time by time generalization matrix shown in Fig.3a showed that the task allowed the 474

formation of stable representations associated with the figures (1 random polygon and 1 rounded shape for AMs 475

and the same for the non-AMs) acting successfully as a localizer for the two types of memories. The bootstrapping 476

analysis performed on a 0–2.5s time-window showed that the classifier could distinguish with a peak accuracy of 477

0.55 between AM and non-AM (p = 0.0129) in a sustained time-window (0 – ~2.2 secs), including a late time 478

window that is most likely related to the representation of the memory itself rather than to any perceptual features 479

of the stimuli. In a second step, the classifier was trained during the presentation of the cue in the retrieval task 480

and then tested on the pain decisiont task in a 1sec time-window starting from the onset of the face. Crucially, 481

any consistency in the neural pattern observed across tasks would show the representation of the memories. The 482



https://doi.org/10.1101/715276



16

bootstrapping analysis revealed a significant cluster (p = 0.032) in a sustained time-window (0.6–1 secs) showing 483

evidence for the online reactivation of the memory in preparation of the empathy judgement with a peak accuracy 484

of 0.53. The result of the classifier across tasks is shown in Fig.3b. 485

486

Fig.3. LDA results a) Sanity check: time by time generalization matrix showing significant classification of AM 487

vs non-Am within the retrieval task. b) Time by time generalization matrix (i.e. training and testing at each time-488

point) showing significant classification of AM vs non-AM across tasks. 489

490

3.3. fMRI results 491

Fig.4 shows masked clusters resulting from the whole-brain analysis. 492

The contrast AM>nonAM for the analysis time-locked to the onset of the context revealed a significant FWE 493

corrected (p <. 05) cluster with a peak in the Precuneus, BA 7, MNI: [3 -64 38], (150 voxels), t(27) = 5.76 p = 494

0.001, in the Superior Parietal Lobule, BA 7, MNI: [-36 -58 59], (114 voxels), t(27) = 4.42, p = 0.003 extending 495

to the Inferior Parietal Lobule (BA 40) and in the Superior Temporal Gyrus (BA 39). This contrast also revealed 496

a cluster with a peak in the Posterior Cingulate, BA 23, MNI: [3 -28 26], (62 voxels), t(27) = 4.71, p = 0.038. 497

Masked clusters showing greater activation for AM as compared to non-AM are depicted in Fig.4a. The opposite 498

contrast did not reveal any significant FWE corrected cluster. 499

Fig.4b shows the result of the contrast AM>nonAM for the analysis time-locked to the onset of the face. Greater 500

activation for AM as compared to non-AM was observed in a significant FWE corrected cluster (p < .05) in the 501

Superior Frontal Gyrus, BA 10, MNI: [-18 62 23], (66 voxels), t(27) = 5.49 p = 0.024, and in a cluster in the 502

Inferior Parietal Lobule, BA 39, MNI: [-36 61 41], (75 voxels), t(27) = 3.67, p = 0.014. This specific region of 503

the IPL is part of the functional fractionation of the TPJ and is considered as part of the core network of the theory 504

of mind (Schurz et al., 2014). The opposite contrast revealed greater activation for non-AM than AM in a 505

significant FWE corrected cluster in the parahippocampal gyrus (PHG), BA 36, MNI: [-18 -16 -22], (169 voxels), 506

t(27) = 6.22, p < 0.001. 507



https://doi.org/10.1101/715276



17

508

Fig. 4. FMRI results Whole-brain analysis results (left panel) and raincloud plots (right panel) of the activation 509

in each condition and each cluster. a) Whole-brain analysis related to the presentation of the context. Only the 510

contrast AM>non-AM showed significant clusters. b) Whole-brain analysis related to the presentation of the face. 511

Figure shows significant clusters resulting from both the AM>non-AM and non-AM>AM contrasts. 512

4. Discussion 513

In the current study we recorded EEG and hemodynamic activity from two independent samples of 514

participants to investigate whether AMs are reactivated in the service of empathy. The present results from two 515

independent experiments provide behavioral, electrophysiological and fMRI evidence in support of a direct 516

engagement of AM reactivation in empathy. We observed direct evidence for AM reactivation when participants 517

were required explicit judgement of their empathy awareness (experiment 1) in an empathy task that activated the 518

brain areas that are critical for both empathy and AM processes (experiment 2). Our experiments show important 519

direct evidence on the role of AM in empathy thus providing insights into the mechanism implied by previous 520

studies suggesting that participants’ past experiences interact with empathic abilities (Bluck et al., 2013; Gaesser 521

and Schacter, 2014; Perry et al., 2011). 522

In experiment 1, EEG was recorded during a task that prompted empathy, and a memory retrieval task. 523

Cortical EEG patterns during the retrieval task were used to probe the data from the empathy task for evidence of 524

reactivation of AMs and non-AMs. We applied an LDA classifier, which was trained and tested across tasks, and 525

found evidence for online memory reactivation when participants explicitly judged their empathy for others’ 526

experiences. Participants could empathize more with people depicted in situations they had experienced 527

themselves as compared to situations that they never experienced, as reflected in self-reported rates of participants’ 528

empathy awareness. This behavioural result was replicated in experiment 2 showing the robustness of this 529

behavioural evidence. 530



https://doi.org/10.1101/715276



18

Three critical features of the study design underwrite the robustness of our findings. First, the 531

autobiographical component of the memories used to probe empathy was unprompted in the empathy task. 532

Participants' AM retrieval could have no impact on the rating of their empathy awareness unless participants based 533

their judgement on their own past experience. Therefore, the EEG evidence for reactivation of AM patterns is 534

remarkable because participants could have relied entirely on their semantic knowledge to perform the tasks, yet 535

the above chance performance of the classifier suggests that they did not. Second, the memory retrieval task was 536

always performed after the empathy task to avoid that participants could be primed to specifically retrieve their 537

own memories. Third, we used perceptually different stimuli to prompt empathy for specific contexts (sentences) 538

and to trigger the reactivation of those episodes (shapes). This was done to avoid any overlap in the perceptual 539

features that cued the memories in the two tasks and ensure that the classifier could only identify the neural 540

fingerprint of the memories reactivation per se. Our results suggest that memory retrieval and empathic processes 541

operate within the same time-frames. The EEG pattern classifier approach has been successfully adopted to 542

differentiate between the retrieval of perceptual and semantic content of an episodic memory (Linde-Domingo et 543

al., 2019) and in different mechanisms of memory (Jafarpour et al., 2013). The timing of the retrieval of an AM 544

has been shown to occur between 400 and 600 ms even when it is only spontaneously recalled (Addante, 2015; 545

Hebscher et al., 2019). The squared shape of the time by time generalization output depicted in Fig.3 shows that 546

the representation is stable across time (King and Dehaene, 2014). Fig.3a shows that the representation of the 547

memories starts stabilizing between 500ms and 1sec and lasts until ~2 secs. Fig.3b shows that the representation 548

of the memories reactivate in the time-window when the empathy judgement was prepared. 549

In experiment 2, we recorded fMRI in an independent sample of participants to verify that the empathy 550

task activated brain areas commonly associated with empathy and AM. This second experiment replicated the 551

behavioural result obtained in experiment 1: participants reported increased empathy awareness for individuals 552

described in contexts for which they had an associated AM. Whole-brain analyses contrasting the hemodynamic 553

response for AM and non-AM were conducted related to the onset of the context, i.e. when participants read the 554

sentences describing either an AM or a non-AM, and of the face, conveying painful or neutral expression. The 555

first analysis showed activation of the Precuneus (BA 7), PCC (BA 23) and left SPL (BA 7). The second analysis 556

activated the left SFG (BA 10) and a specific region of the left IPL, part of the functional fractionation of the TPJ 557

(BA 39). The activation of these brain areas is consistent with previous literature showing that these brain areas 558

underlie both AM retrieval and empathic processes (Amodio & Frith, 2006; Bernhardt & Singer, 2012; Buckner 559

& Carroll, 2007; Frith & Frith, 2003; Spreng et al., 2008; Zaki & Ochsner, 2012). 560

The parietal cortex is a critical hub for cognitive, cold, empathic processes, and AM retrieval. The SPL 561

is involved in the online maintenance of relevant information (Postle et al., 2004; Xie et al., 2019) and in the 562

retrieval of specific AM (Addis et al., 2004). A recent study by Hebscher and colleagues, demonstrated the causal 563

involvement of the precuneus in AM retrieval (Hebscher et al., 2020). The involvement of the parietal cortex in 564

the retrieval of AM, and in particular of the precuneus, has been suggested to be responsible of the spontaneous 565

AM retrieval from an egocentric perspective (Freton et al., 2014) and in flexible perspective shifting during AM 566

retrieval (St. Jacques et al., 2017). Therefore, it ultimately contributes to the vividness of the retrieval and of 567

constructing realistic mental images (Fuentemilla et al., 2014). The precuneus is reliably engaged in the network 568

of brain areas underlying the understanding of others’ mind, i.e. cold empathy (Molenberghs et al., 2016). The 569



https://doi.org/10.1101/715276



19

PCC is involved in the retrieval of familiar objects and places (Burianova and Grady, 2007) and together with at 570

least the anterior division of the precuneus, in self-referential processes (Sajonz et al., 2010). 571

In the whole-brain analysis contrasting the AM and non-AM related to the onset of the face, we observed 572

the activation of a specific region of the left IPL, part of the functional fractionation of the TPJ (BA 39) and of 573

the left SFG (BA 10). A recent meta-analysis investigating the core network of theory of mind (Schurz et al., 574

2014) demonstrated that, together with the mPFC, the left TPJ is a core brain area of this network (Gaesser et al., 575

2019). Lesion studies further support this view as damage of the left TPJ can selectively reduce theory of mind 576

abilities but not other cognitive or executive abilities (Apperly et al., 2004; Bzdok et al., 2013; Samson et al., 577

2004). The activation of the left SFG was in line with the source estimation of the ERP data in experiment 1 for 578

the same contrast and time-window. ERP studies investigating empathy for physical pain have shown that an 579

empathic reaction, reflecting the processing of a painful experience, is expressed as a positive shift of the ERP 580

response, compared to a neutral condition with (e.g., Meconi, Doro, Lomoriello, Mastrella, & Sessa, 2018; Sessa 581

& Meconi, 2015) or without (Sheng and Han, 2012) relation to explicit or implicit measures of empathy. In 582

experiment 1, we observed a positive shift in the ERPs reflecting the processing of painful when compared to 583

neutral faces within 1 second in a cluster of centro-parietal electrodes that was estimated to be generated in the 584

IPL and the IFG. Within the same time-window, ERPs time-locked to the onset of the faces reflecting the 585

processing of the preceding memory showed a positive shift of the ERPs for AM as compared to non-AM. The 586

neural source of this effect was estimated to be in the SFG. According to the multiple memory system of social 587

cognition (MMS), prejudice and stereotyping are the result of affective and semantic associations in memory 588

(Amodio and Ratner, 2011) resulting from autobiographical experience as well as from acquired knowledge. 589

Studies on cross-racial empathy for pain showed that empathic responses are more natural for own-race faces or 590

more familiar faces when compared to other-race faces (Avenanti et al., 2010; Sessa et al., 2014a; Xu et al., 2009). 591

These ERP studies therefore provided some parallel evidence that past experiences and shared cultural background 592

can influence empathy as they contribute to reduce the psychological distance between the observer and the target 593

of empathy (Meconi et al., 2015). Our ERP results and the source analysis regarding the face and memory effects 594

replicate previous ERP (Sessa et al., 2014b) neuroimaging studies on the neural correlates of empathy (Bernhardt 595

& Singer, 2012; Fan, Duncan, de Greck, & Northoff, 2011; Lamm, Decety, & Singer, 2011) and are in line with 596

our fMRI results obtained in experiment 2. However, it is important to mention that the area obtained from the 597

source reconstruction in experiment 1, i.e. the SFG, seems not to fully overlap with the one observed in experiment 598

2 from the fMRI. It is possible that the wider cluster obtained in experiment 1 could have been due to lower 599

precision in the source reconstruction of an ERP effect compared with source localization from the fMRI data. 600

Equally, though it is possible that the MNI coordinates in the fMRI analysis only identify the peaks of the clusters 601

that in fact reflect the activation of the same, larger, brain area identified from the ERP data. 602

In the current study, we did not observe the activation of the MTL in the contrast AM>non-AM. Notably, 603

the participants’ AMs were on average 5 years old. A recent review (Barry and Maguire, 2019) highlighted that 604

although memories seem to become independent from hippocampal activation with remoteness in time, the 605

hippocampus remains involved in context/memory reconstruction (Zeidman and Maguire, 2016) even though the 606

original memory trace is with time transferred to the neocortex. We did observe the activation of the PHG in the 607

contrast non-AM>AM. This result was surprising and we speculate that it is consistent with the mindreading 608

hypothesis (Gaesser, 2018) that draws on those studies with healthy participants showing the involvement of the 609



https://doi.org/10.1101/715276



20

episodic simulation in performing tasks that prompt cold empathy. Consistently, patients with medial temporal 610

lobe lesions do not show increases in empathy when prompted to use episodic simulation to construct specific 611

episodes of others suffering (Sawczak et al., 2019). 612

5. Conclusions 613

The present study provides important evidence of a re-activation of AMs in the context of empathy. However, 614

puzzling previous evidence showing little empathy impairment in patients with amnesia opens future research 615

question on whether memory causally drives empathy judgments. This would require future work that modulates 616

memory retrieval in a time-sensitive manner. 617

618

619



https://doi.org/10.1101/715276



21

Acknowledgements. 620

The Authors thank all the participants and Ross Wilson and Nina Salman for helping with fMRI data collection. 621

Funding: This work was supported by the European Union's Horizon 2020, MSCA-IF-2015 (Nº702530), and by 622

the ESRC (NºES/S001964/1) awarded to F.M. S.H. is supported by grants from the ERC (Nº647954), the ESRC 623

(NºES/R010072/1), and the Wolfson Society and Royal Society. 624

625

626

627



https://doi.org/10.1101/715276



22

References 628

Addante, R.J., 2015. A critical role of the human hippocampus in an electrophysiological measure of 629

implicit memory. NeuroImage 109, 515. https://doi.org/10.1016/j.neuroimage.2014.12.069 630

Addis, D.R., McIntosh, A.R., Moscovitch, M., Crawley, A.P., McAndrews, M.P., 2004. 631

Characterizing spatial and temporal features of autobiographical memory retrieval networks: 632

a partial least squares approach. NeuroImage 23, 1460–1471. 633

https://doi.org/10.1016/j.neuroimage.2004.08.007 634

Amodio, D.M., Frith, C.D., 2006. Meeting of minds: the medial frontal cortex and social cognition. 635

Nature Reviews Neuroscience 7, 268–277. https://doi.org/10.1038/nrn1884 636

Amodio, D.M., Ratner, K.G., 2011. A Memory Systems Model of Implicit Social Cognition. Curr Dir 637

Psychol Sci 20, 143–148. https://doi.org/10.1177/0963721411408562 638

Antony, J.W., Ferreira, C.S., Norman, K.A., Wimber, M., 2017. Retrieval as a Fast Route to Memory 639

Consolidation. Trends in Cognitive Sciences 21, 573–576. 640

https://doi.org/10.1016/j.tics.2017.05.001 641

Apperly, I.A., Samson, D., Chiavarino, C., Humphreys, G.W., 2004. Frontal and Temporo-Parietal 642

Lobe Contributions to Theory of Mind: Neuropsychological Evidence from a False-Belief 643

Task with Reduced Language and Executive Demands. Journal of Cognitive Neuroscience 644

16, 1773–1784. https://doi.org/10.1162/0898929042947928 645

Avenanti, A., Sirigu, A., Aglioti, S.M., 2010. Racial Bias Reduces Empathic Sensorimotor Resonance 646

with Other-Race Pain. Current Biology 20, 1018–1022. 647

https://doi.org/10.1016/j.cub.2010.03.071 648

Bagby, R.M., Parker, J.D.A., Taylor, G.J., 1994. The twenty-item Toronto Alexithymia scale—I. Item 649

selection and cross-validation of the factor structure. Journal of Psychosomatic Research 38, 650

23–32. https://doi.org/10.1016/0022-3999(94)90005-1 651

Baron-Cohen, S., Wheelwright, S., 2004. The Empathy Quotient: An Investigation of Adults with 652

Asperger Syndrome or High Functioning Autism, and Normal Sex Differences. J Autism Dev 653

Disord 34, 163–175. https://doi.org/10.1023/B:JADD.0000022607.19833.00 654

Barry, D.N., Maguire, E.A., 2019. Remote Memory and the Hippocampus: A Constructive Critique. 655

Trends in Cognitive Sciences 23, 128–142. https://doi.org/10.1016/j.tics.2018.11.005 656

Beadle, J.N., Tranel, D., Cohen, N.J., Duff, M., 2013. Empathy in Hippocampal Amnesia. Front. 657

Psychol. 4. https://doi.org/10.3389/fpsyg.2013.00069 658

Bernhardt, B.C., Singer, T., 2012a. The Neural Basis of Empathy. Annual Review of Neuroscience 659

35, 1–23. https://doi.org/10.1146/annurev-neuro-062111-150536 660

Bernhardt, B.C., Singer, T., 2012b. The Neural Basis of Empathy. Annual Review of Neuroscience 661

35, 1–23. https://doi.org/10.1146/annurev-neuro-062111-150536 662



https://doi.org/10.1101/715276



23

Blankertz, B., Lemm, S., Treder, M., Haufe, S., Müller, K.-R., 2011. Single-trial analysis and 663

classification of ERP components — A tutorial. NeuroImage, Multivariate Decoding and 664

Brain Reading 56, 814–825. https://doi.org/10.1016/j.neuroimage.2010.06.048 665

Bluck, S., Baron, J.M., Ainsworth, S.A., Gesselman, A.N., Gold, K.L., 2013. Eliciting Empathy for 666

Adults in Chronic Pain through Autobiographical Memory Sharing. Applied Cognitive 667

Psychology 27, 81–90. https://doi.org/10.1002/acp.2875 668

Boccia, M., Teghil, A., Guariglia, C., 2019. Looking into recent and remote past: Meta-analytic 669

evidence for cortical re-organization of episodic autobiographical memories. Neuroscience & 670

Biobehavioral Reviews 107, 84–95. https://doi.org/10.1016/j.neubiorev.2019.09.003 671

Buckner, R.L., Carroll, D.C., 2007. Self-projection and the brain. Trends in Cognitive Sciences 11, 672

49–57. https://doi.org/10.1016/j.tics.2006.11.004 673

Burianova, H., Grady, C.L., 2007. Common and unique neural activations in autobiographical, 674

episodic, and semantic retrieval. Journal of cognitive neuroscience 19, 1520–1534. 675

Bzdok, D., Langner, R., Schilbach, L., Jakobs, O., Roski, C., Caspers, S., Laird, A.R., Fox, P.T., 676

Zilles, K., Eickhoff, S.B., 2013. Characterization of the temporo-parietal junction by 677

combining data-driven parcellation, complementary connectivity analyses, and functional 678

decoding. Neuroimage 81, 381–392. https://doi.org/10.1016/j.neuroimage.2013.05.046 679

Cabeza, R., St Jacques, P., 2007. Functional neuroimaging of autobiographical memory. Trends in 680

Cognitive Sciences 11, 219–227. https://doi.org/10.1016/j.tics.2007.02.005 681

Chen, Y.-C., Chen, C.-C., Decety, J., Cheng, Y., 2014. Aging is associated with changes in the neural 682

circuits underlying empathy. Neurobiology of Aging 35, 827–836. 683

https://doi.org/10.1016/j.neurobiolaging.2013.10.080 684

Corcoran, R., Frith, C.D., 2003. Autobiographical memory and theory of mind: evidence of a 685

relationship in schizophrenia. Psychological Medicine 33, 897–905. 686

https://doi.org/10.1017/S0033291703007529 687

Cotelli, M., Calabria, M., Manenti, R., Rosini, S., Maioli, C., Zanetti, O., Miniussi, C., 2012. Brain 688

stimulation improves associative memory in an individual with amnestic mild cognitive 689

impairment. Neurocase 18, 217–223. https://doi.org/10.1080/13554794.2011.588176 690

Danziger, N., Faillenot, I., Peyron, R., 2009. Can We Share a Pain We Never Felt? Neural Correlates 691

of Empathy in Patients with Congenital Insensitivity to Pain. Neuron 61, 203–212. 692

https://doi.org/10.1016/j.neuron.2008.11.023 693

Davis, M.H., 1983. Measuring individual differences in empathy: Evidence for a multidimensional 694

approach. Journal of Personality and Social Psychology 44, 113–126. 695

https://doi.org/10.1037/0022-3514.44.1.113 696

Drost, R., Postma, A., Oudman, E., 2019. Cognitive and affective theory of mind in Korsakoff’s 697

syndrome. Acta Neuropsychiatrica 31, 128–134. https://doi.org/10.1017/neu.2018.35 698



https://doi.org/10.1101/715276



24

Duval, C., Bejanin, A., Piolino, P., Laisney, M., de La Sayette, V., Belliard, S., Eustache, F., 699

Desgranges, B., 2012. Theory of mind impairments in patients with semantic dementia. Brain 700

135, 228–241. https://doi.org/10.1093/brain/awr309 701

Duval, C., Piolino, P., Bejanin, A., Eustache, F., Desgranges, B., 2011. Age effects on different 702

components of theory of mind. Conscious Cogn 20, 627–642. 703

https://doi.org/10.1016/j.concog.2010.10.025 704

Fairhurst, M., Fairhurst, K., Berna, C., Tracey, I., 2012. An fMRI study exploring the overlap and 705

differences between neural representations of physical and recalled pain., An fMRI Study 706

Exploring the Overlap and Differences between Neural Representations of Physical and 707

Recalled Pain. PLoS One 7, 7, e48711–e48711. 708

https://doi.org/10.1371/journal.pone.0048711, 10.1371/journal.pone.0048711 709

Fan, Y., Duncan, N.W., de Greck, M., Northoff, G., 2011. Is there a core neural network in empathy? 710

An fMRI based quantitative meta-analysis. Neuroscience & Biobehavioral Reviews 35, 903–711

911. https://doi.org/10.1016/j.neubiorev.2010.10.009 712

Forkmann, K., Wiech, K., Sommer, T., Bingel, U., 2015. Reinstatement of pain-related brain 713

activation during the recognition of neutral images previously paired with nociceptive stimuli: 714

PAIN 156, 1501–1510. https://doi.org/10.1097/j.pain.0000000000000194 715

Freton, M., Lemogne, C., Bergouignan, L., Delaveau, P., Lehéricy, S., Fossati, P., 2014. The eye of 716

the self: precuneus volume and visual perspective during autobiographical memory retrieval. 717

Brain Structure and Function 219, 959–968. 718

Frith, U., Frith, C.D., 2003. Development and neurophysiology of mentalizing. Philos. Trans. R. Soc. 719

Lond., B, Biol. Sci. 358, 459–473. https://doi.org/10.1098/rstb.2002.1218 720

Fuentemilla, L., Barnes, G.R., Düzel, E., Levine, B., 2014. Theta oscillations orchestrate medial 721

temporal lobe and neocortex in remembering autobiographical memories. Neuroimage 85 Pt 722

2, 730–737. https://doi.org/10.1016/j.neuroimage.2013.08.029 723

Gaesser, B., 2018. Episodic Mind Reading: Mentalizing Guided by Imagined and Remembered 724

Scenes. https://doi.org/10.31234/osf.io/acv9r 725

Gaesser, B., Hirschfeld-Kroen, J., Wasserman, E.A., Horn, M., Young, L., 2019. A role for the medial 726

temporal lobe subsystem in guiding prosociality: the effect of episodic processes on 727

willingness to help others. Soc Cogn Affect Neurosci 14, 397–410. 728

https://doi.org/10.1093/scan/nsz014 729

Gaesser, B., Schacter, D.L., 2014. Episodic simulation and episodic memory can increase intentions 730

to help others. PNAS 111, 4415–4420. https://doi.org/10.1073/pnas.1402461111 731

Hebscher, M., Ibrahim, C., Gilboa, A., 2020. Precuneus stimulation alters the neural dynamics of 732

autobiographical memory retrieval. Neuroimage 210, 116575. 733

Hebscher, M., Ibrahim, C., Gilboa, A., 2019. Precuneus stimulation alters the neural dynamics of 734

autobiographical memory retrieval. bioRxiv 592436. https://doi.org/10.1101/592436 735



https://doi.org/10.1101/715276



25

Jafarpour, A., Horner, A.J., Fuentemilla, L., Penny, W.D., Duzel, E., 2013. Decoding oscillatory 736

representations and mechanisms in memory. Neuropsychologia 51, 772–780. 737

https://doi.org/10.1016/j.neuropsychologia.2012.04.002 738

King, J.-R., Dehaene, S., 2014. Characterizing the dynamics of mental representations: the temporal 739

generalization method. Trends Cogn Sci 18, 203–210. 740

https://doi.org/10.1016/j.tics.2014.01.002 741

Lamm, C., Decety, J., Singer, T., 2011. Meta-analytic evidence for common and distinct neural 742

networks associated with directly experienced pain and empathy for pain. NeuroImage 54, 743

2492–2502. https://doi.org/10.1016/j.neuroimage.2010.10.014 744

Linde-Domingo, J., Treder, M.S., Kerrén, C., Wimber, M., 2019. Evidence that neural information 745

flow is reversed between object perception and object reconstruction from memory. Nature 746

Communications 10, 179. https://doi.org/10.1038/s41467-018-08080-2 747

Macey, P.M., Macey, K.E., Kumar, R., Harper, R.M., 2004. A method for removal of global effects 748

from fMRI time series. NeuroImage 22, 360–366. 749


McClelland, J.L., McNaughton, B.L., O’Reilly, R.C., 1995. Why there are complementary learning 751

systems in the hippocampus and neocortex: Insights from the successes and failures of 752

connectionist models of learning and memory. Psychological Review 102, 419–457. 753

https://doi.org/10.1037/0033-295X.102.3.419 754

Meconi, F., Anderl-Straub, S., Raum, H., Landgrebe, M., Langguth, B., Bäuml, K.-H.T., Hanslmayr, 755

S., 2016. Aberrant prefrontal beta oscillations predict episodic memory encoding deficits in 756

schizophrenia. NeuroImage: Clinical 12, 499–505. https://doi.org/10.1016/j.nicl.2016.08.017 757

Meconi, F., Doro, M., Lomoriello, A.S., Mastrella, G., Sessa, P., 2018. Neural measures of the role of 758

affective prosody in empathy for pain. Scientific Reports 8, 291. 759

https://doi.org/10.1038/s41598-017-18552-y 760

Meconi, F., Vaes, J., Sessa, P., 2015. On the neglected role of stereotypes in empathy toward other-761

race pain. Social Neuroscience 10, 1–6. https://doi.org/10.1080/17470919.2014.954731 762

Michelmann, S., Bowman, H., Hanslmayr, S., 2016. The Temporal Signature of Memories: 763

Identification of a General Mechanism for Dynamic Memory Replay in Humans. PLOS 764

Biology 14, e1002528. https://doi.org/10.1371/journal.pbio.1002528 765

Molenberghs, P., Johnson, H., Henry, J.D., Mattingley, J.B., 2016. Understanding the minds of others: 766

A neuroimaging meta-analysis. Neuroscience & Biobehavioral Reviews 65, 276–291. 767

Monetta, L., Grindrod, C.M., Pell, M.D., 2009. Irony comprehension and theory of mind deficits in 768

patients with Parkinson’s disease. Cortex 45, 972–981. 769

https://doi.org/10.1016/j.cortex.2009.02.021 770

Moreau, N., Rauzy, S., Bonnefoi, B., Renié, L., Martinez-Almoyna, L., Viallet, F., Champagne-771

Lavau, M., 2015. Different Patterns of Theory of Mind Impairment in Mild Cognitive 772



https://doi.org/10.1101/715276



26

Impairment. Journal of Alzheimer’s Disease 45, 581–597. https://doi.org/10.3233/JAD-773

143021 774

Moreau, N., Viallet, F., Champagne-Lavau, M., 2013. Using memories to understand others: The role 775

of episodic memory in theory of mind impairment in Alzheimer disease. Ageing Research 776

Reviews 12, 833–839. https://doi.org/10.1016/j.arr.2013.06.005 777

Oosterman, J.M., Derksen, L.C., van Wijck, A.J.M., Veldhuijzen, D.S., Kessels, R.P.C., 2011. 778

Memory Functions in Chronic Pain: Examining Contributions of Attention and Age to Test 779

Performance. The Clinical Journal of Pain 27, 70. 780

https://doi.org/10.1097/AJP.0b013e3181f15cf5 781

Pell, M.D., Monetta, L., Rothermich, K., Kotz, S.A., Cheang, H.S., McDonald, S., 2014. Social 782

perception in adults with Parkinson’s disease. Neuropsychology 28, 905–916. 783

https://doi.org/10.1037/neu0000090 784

Perry, D., Hendler, T., Shamay-Tsoory, S.G., 2011. Projecting memories: The role of the 785

hippocampus in emotional mentalizing. NeuroImage 54, 1669–1676. 786


Postle, B.R., Awh, E., Jonides, J., Smith, E.E., D’Esposito, M., 2004. The where and how of 788

attention-based rehearsal in spatial working memory. Cognitive Brain Research 20, 194–205. 789

Rabin, J.S., Rosenbaum, R.S., 2012. Familiarity modulates the functional relationship between theory 790

of mind and autobiographical memory. NeuroImage 62, 520–529. 791


Ramanan, S., de Souza, L.C., Moreau, N., Sarazin, M., Teixeira, A.L., Allen, Z., Guimarães, H.C., 793

Caramelli, P., Dubois, B., Hornberger, M., Bertoux, M., 2017. Determinants of theory of 794

mind performance in Alzheimer’s disease: A data-mining study. Cortex 88, 8–18. 795

https://doi.org/10.1016/j.cortex.2016.11.014 796

Rosenbaum, R.S., Stuss, D.T., Levine, B., Tulving, E., 2007. Theory of Mind Is Independent of 797

Episodic Memory. Science 318, 1257–1257. https://doi.org/10.1126/science.1148763 798

Rubin, D.C., Schrauf, R.W., Greenberg, D.L., 2003. Belief and recollection of autobiographical 799

memories. Memory & Cognition 31, 887–901. https://doi.org/10.3758/BF03196443 800

Sajonz, B., Kahnt, T., Margulies, D.S., Park, S.Q., Wittmann, A., Stoy, M., Ströhle, A., Heinz, A., 801

Northoff, G., Bermpohl, F., 2010. Delineating self-referential processing from episodic 802

memory retrieval: common and dissociable networks. Neuroimage 50, 1606–1617. 803

Samson, D., Apperly, I.A., Chiavarino, C., Humphreys, G.W., 2004. Left temporoparietal junction is 804

necessary for representing someone else’s belief. Nature Neuroscience 7, 499. 805

https://doi.org/10.1038/nn1223 806

Sawczak, C., McAndrews, M.P., Gaesser, B., Moscovitch, M., 2019. Episodic simulation and 807

empathy in older adults and patients with unilateral medial temporal lobe excisions. 808

Neuropsychologia 135, 107243. https://doi.org/10.1016/j.neuropsychologia.2019.107243 809



https://doi.org/10.1101/715276



27

Schurz, M., Radua, J., Aichhorn, M., Richlan, F., Perner, J., 2014. Fractionating theory of mind: A 810

meta-analysis of functional brain imaging studies. Neuroscience & Biobehavioral Reviews 811

42, 9–34. https://doi.org/10.1016/j.neubiorev.2014.01.009 812

Sessa, P., Meconi, F., 2015. Perceived trustworthiness shapes neural empathic responses toward 813

others’ pain. Neuropsychologia 79, 97–105. 814

https://doi.org/10.1016/j.neuropsychologia.2015.10.028 815

Sessa, P., Meconi, F., Castelli, L., Dell’Acqua, R., 2014a. Taking one’s time in feeling other-race 816

pain: an event-related potential investigation on the time-course of cross-racial empathy. Soc 817

Cogn Affect Neurosci 9, 454–463. https://doi.org/10.1093/scan/nst003 818

Sessa, P., Meconi, F., Han, S., 2014b. Double dissociation of neural responses supporting perceptual 819

and cognitive components of social cognition: Evidence from processing of others’ pain. 820

Scientific Reports 4, 7424. https://doi.org/10.1038/srep07424 821

Shamay-Tsoory, S.G., Aharon-Peretz, J., Perry, D., 2009. Two systems for empathy: a double 822

dissociation between emotional and cognitive empathy in inferior frontal gyrus versus 823

ventromedial prefrontal lesions. Brain 132, 617–627. https://doi.org/10.1093/brain/awn279 824

Sheng, F., Han, S., 2012. Manipulations of cognitive strategies and intergroup relationships reduce the 825

racial bias in empathic neural responses. NeuroImage 61, 786–797. 826


Spreng, R.N., Grady, C.L., 2009. Patterns of Brain Activity Supporting Autobiographical Memory, 828

Prospection, and Theory of Mind, and Their Relationship to the Default Mode Network. 829

Journal of Cognitive Neuroscience 22, 1112–1123. https://doi.org/10.1162/jocn.2009.21282 830

Spreng, R.N., Mar, R.A., Kim, A.S.N., 2008. The Common Neural Basis of Autobiographical 831

Memory, Prospection, Navigation, Theory of Mind, and the Default Mode: A Quantitative 832

Meta-analysis. Journal of Cognitive Neuroscience 21, 489–510. 833

https://doi.org/10.1162/jocn.2008.21029 834

St. Jacques, P.L., Szpunar, K.K., Schacter, D.L., 2017. Shifting visual perspective during retrieval 835

shapes autobiographical memories. NeuroImage 148, 103–114. 836

Staniloiu, A., Borsutzky, S., Woermann, F., Markowitsch, H.J., 2013. Social cognition in a case of 837

amnesia with neurodevelopmental mechanisms. Front. Psychol. 4. 838

https://doi.org/10.3389/fpsyg.2013.00342 839

Stelzer, J., Chen, Y., Turner, R., 2013. Statistical inference and multiple testing correction in 840

classification-based multi-voxel pattern analysis (MVPA): Random permutations and cluster 841

size control. NeuroImage 65, 69–82. https://doi.org/10.1016/j.neuroimage.2012.09.063 842

Van Veen, B.D., Drongelen, W.V., Yuchtman, M., Suzuki, A., 1997. Localization of brain electrical 843

activity via linearly constrained minimum variance spatial filtering. IEEE Transactions on 844

Biomedical Engineering 44, 867–880. https://doi.org/10.1109/10.623056 845



https://doi.org/10.1101/715276



28

Willenbockel, V., Sadr, J., Fiset, D., Horne, G.O., Gosselin, F., Tanaka, J.W., 2010. Controlling low-846

level image properties: The SHINE toolbox. Behavior Research Methods 42, 671–684. 847

https://doi.org/10.3758/BRM.42.3.671 848

Xi, C., Zhu, Y., Mu, Y., Chen, B., Dong, B., Cheng, H., Hu, P., Zhu, C., Wang, K., 2015. Theory of 849

mind and decision-making processes are impaired in Parkinson’s disease. Behavioural Brain 850

Research 279, 226–233. https://doi.org/10.1016/j.bbr.2014.11.035 851

Xie, Y.J., Li, Y.Y., Xie, B., Xu, Y.Y., Peng, L., 2019. The neural basis of complex audiovisual 852

objects maintenances in working memory. Neuropsychologia 133, 107189. 853

https://doi.org/10.1016/j.neuropsychologia.2019.107189 854

Xu, X., Zuo, X., Wang, X., Han, S., 2009. Do You Feel My Pain? Racial Group Membership 855

Modulates Empathic Neural Responses. J. Neurosci. 29, 8525–8529. 856

https://doi.org/10.1523/JNEUROSCI.2418-09.2009 857

Zaki, J., Ochsner, K.N., 2012. The neuroscience of empathy: progress, pitfalls and promise. Nat 858

Neurosci 15, 675–680. https://doi.org/10.1038/nn.3085 859

Ze, O., Thoma, P., Suchan, B., 2014. Cognitive and affective empathy in younger and older 860

individuals. Aging & Mental Health 18, 929–935. 861

https://doi.org/10.1080/13607863.2014.899973 862

Zeidman, P., Maguire, E.A., 2016. Anterior hippocampus: the anatomy of perception, imagination and 863

episodic memory. Nature Reviews Neuroscience 17, 173–182. 864

https://doi.org/10.1038/nrn.2015.24 865

866

867



https://doi.org/10.1101/715276



29

Figure captions 868

Fig.1. a) Schematic representation of the empathy task used in experiment1 and experiment2. Participants were 869

required to rate how much empathy they felt for the person depicted in the preceding context. b) Schematic 870

representation of the retrieval task used in experiment1 that was used to train the LDA classifier. Participants first 871

learnt to associate four abstract figures with the same sentences describing painful contexts presented during the 872

empathy task (not shown here). In the actual task, for each trial participants were presented with one of the four 873

figures and had to picture in their mind’s eye the context that they learnt to associate with that specific figure. c) 874

Raincloud plots of the subjective reports of participants’ empathy awareness in experiment1 and experiment2. d) 875

Concept of the study; when we encounter someone who shares our same physically painful experience, memory 876

of that experience is reactivated to empathize. 877

Fig.2. Left panel: ERPs time-locked to the onset of the face and reflecting AM and non-AM at the frontal and the 878

posterior cluster. Right top panel: clusters analysis performed over all the electrodes in a 0 – 1 sec time-window. 879

Colors code t-values. Right bottom panel: source localization of the AM vs non-AM contrast. 880

Fig.3. a) Sanity check: time by time generalization matrix showing significant classification of AM vs non-Am 881

within the retrieval task. b) Time by time generalization matrix (i.e. training and testing at each time-point) 882

showing significant classification of AM vs non-AM across tasks. 883

Fig.4. Whole-brain analysis results (left panel) and raincloud plots (right panel) of the activation in each condition 884

and each cluster. A) Whole-brain analysis related to the presentation of the context. Only the contrast AM>non-885

AM showed significant clusters. B) Whole-brain analysis related to the presentation of the face. Figure shows 886

significant clusters resulting from both the AM>non-AM and non-AM>AM contrasts. 887

888



https://doi.org/10.1101/715276



30

Table caption 889

Table 1. Participants’ scores to the Interpersonal Reactivity Index in both experiments. T-tests are performed on 890

the two independent samples. 891

892



https://doi.org/10.1101/715276



31

Author contribution. 893

F.M. formulated the research question, collected and analysed all the data, manually drew the ROIs around the 894

hippocampus and wrote the manuscript. F.M., S.H. and I.A.A. designed the studies. S.H. and I.A.A. supervised 895

the analysis and substantially contributed to the writing of the manuscript. C.F. helped with the analysis of the 896

fMRI data. B.S. supervised the fMRI analysis and the drawing of the ROIs. J.L.D. and S.M. helped with the 897

classifier and bootstrapping analysis. All the authors gave important feedback and comments to the manuscript. 898

Data sharing statement 899

The study was not formally pre-registered but the data from this research are available to view in the OSF 900

repository: https://osf.io/9z2uf/?view_only=6b4f7e6d52a3411bbb6b4343aff79607 901

Conflict of interest. 902

The authors declare no competing financial interests. 903

904



https://doi.org/10.1101/715276


autobiographical memory reactivation in empathy: eeg and ...autobiographical memory reactivation in...

Documents